|
|
|
The Secret of the West ( Le Secret de l'Occident)
quoted by Philipp Hoffman in California in an on-line paper
about the causes of the Western miracle
( shortcut 1, shortcut 2).
PDF-version of article.
Unfortunately, P. Hoffman misquotes this book,
being mistaken about its philosophy. Besides, he does
not realize that his own idea, and much more, is already encompassed
in The Secret of the West.
(Philipp T. Hoffman: "Why Is It That Europeans Ended Up Conquering the Rest of the Globe?
Prices, the Military Revolution, and Western Europe’s Comparative Advantage in Violence",
23 October 2006)
Safety copy: August 2007.
Source
|
|
Why Is It That Europeans Ended Up Conquering the Rest of the Globe? Prices, the Military Revolution, and Western Europe’s Comparative Advantage in Violence
1
Why Is It That Europeans Ended Up Conquering the Rest of the Globe?
Prices, the Military Revolution, and Western Europe’s Comparative Advantage in Violence
Philip T. Hoffman
pth@hss.caltech.edu
California Institute of Technology
HSS 228-77 Pasadena, CA 91125 USA
October 23, 2006
Abstract
Why did Europeans conquer the rest of the world? The likely cause was a tournament among
western European rulers that fostered military innovation. Price data from England, France, and
Germany support such an argument, as do physical measures of military productivity; they show
that the military sector in western Europe was experiencing rapid and sustained technical change
well before the Industrial Revolution. The price data shed new light on this military revolution
and its economic consequences. Comparisons with the rest the world explain why it was
peculiar to Europe and why it gave western Europe a comparative advantage in violence.
2
In recent years, historians, economists, and other social scientists have
energetically debated when Western Europe first forged ahead of other parts of the
world–in particular, advanced parts of Asia–in the race toward economic development.
Was it only after 1800, with the Industrial Revolution well underway, that Western
European per-capita incomes, labor productivity, or technology diverged (Wong 1997;
Pomeranz 2000; Goldstone forthcoming)? Or was it earlier, before the Industrial
Revolution (van Zanden 2003; Allen 2005; Broadberry and Gupta 2005)? And what was
the cause of the divergence? Was it beneficial institutions, which encouraged investment
and the accumulation of human and physical capital (North and Thomas 1973; North and
Weingast 1989; Acemoglu, Johnson et al. 2002)? The Scientific Revolution and the
Enlightenment, which spread useful knowledge and political reform (Jacob 1997; Mokyr
2002;
Cosandey 1997)?
Or was it simply an accident that the Industrial Revolution
started in Western Europe (Clark 2003)?
In this debate, one area in which Western Europe possessed an undeniable
comparative advantage well before 1800 seems to have been overlooked–namely,
violence. The states of Western Europe were simply better at making and using artillery,
firearms, fortifications, and armed ships than other advanced parts of the world and they
had this advantage long before 1800. By 1800, Europeans had conquered some 35
percent of the globe, and they controlled lucrative trade routes as far away as Asia
(Parker 1996, 5). Some of the land they subjugated had come into their hands because of
new diseases that they introduced into vulnerable populations, and in these instances–in
the Americas in particular–their advantage was not military, but biological (Diamond
1997). But other inhabitants of densely populated parts of Eurasia would have had the
same biological edge. Why was it therefore the Western Europeans who took over the
Americas, and not the Chinese or the Japanese?
The history of conquest is not the only evidence for Western Europe’s military
advantage before 1800. States elsewhere–China, Japan, and the Ottoman Empire–
certainly possessed firearms or ships equipped with artillery, but by the late seventeenth
century, if not beforehand, nearly all of them had fallen behind in using this technology.
The case of the Ottoman Empire is illustrative. There the military gap may reach back as
far as 1572, when Venetian cannon founders judged that guns captured during the naval
battle at Lepanto were simply not worth reusing. The Ottoman cannons had to be melted
down–and new metal had to be added to the mixture–because “the material is of such
poor quality.” (Mallett and Hale 1984, 400).
At a time when the high cost of
manufactured goods meant everything was salvaged—even clothing from fallen
comrades—that amounts to strong evidence from revealed preference about how much
better Western European weapons had become. The history of trade and of the migration
of military experts points in the same direction. Although the Ottomans could threaten
Vienna as late as 1683, they were importing weapons from western Europe and often
relied on the expertise of European military specialists.
1
The Ottoman Empire was hardly exceptional. From the Middle East to East Asia,
experts from Western Europe were hired in Asia to provide needed help with gun
making, tactics, and military organization. They ranged from renegade European gun
founders in the sixteenth century to Napoleonic officers the early 1800s. In seventeenth-
century China, even Jesuit missionaries were pressed into service to help the Chinese
Emperor make better cannons. The evidence for Western Europe’s military prowess is so
3
strong that it has even convinced some of the historians who argue against any
divergence between Western Europe and advanced areas of China before 1800.
Although they would argue that Western Europe was not wealthier or more developed
than rich areas of China, they would acknowledge that its military technology was more
advanced (Wong 1997, 89-90; Pomeranz 2000, 199-200).
The evidence is thus fairly clear, but it is nonetheless surprising that western
Europe had come to dominate this technology of gunpowder weapons so early. Firearms
and gunpowder, after all, had originated in China and spread throughout Eurasia. States
outside Western Europe possessed the revolutionary weapons and did become, at least for
a while, proficient at manufacturing or exploiting the new military technology. The
Ottomans, for instance, made high quality artillery as late as the 1500s. The Japanese
independently discovered, at about the same time as Western Europeans, the key tactical
innovation (volley fire) that allowed infantry soldiers with slow loading muskets to
maintain a nearly continuous round of fire.
2
Yet by the late seventeenth century, if not
before, Chinese, Japanese, and Ottoman military technology and tactics all lagged far
behind what one found in western Europe.
Why did these other powerful states fall behind? Apart from Carlo Cipolla’s
(1966) pioneering effort some 40 years ago, economist historians (and social scientists in
general) have not paid much attention to this question. Western Europe’s advances in
military tactics and technology have certainly attracted a number of talented military
historians and historians of technology, but their work ignores the economics, even
though they acknowledge that the cost of weapons fell.
4
What happens if we examine the
political economy of the military revolution and look in particular at prices of military
goods? What do they tell us about western Europe’s military growing military strength?
The price data, it turns out, offer some novel insights into the debates military and
technological historians have had over the nature of the military revolution. They also
carry the startling implication that Europe’s military sector could sustain technical change
for centuries–a feat virtually unknown elsewhere in pre-industrial economies. But their
greatest signifance lies with what they suggest is the underlying cause of Western
Europe’s comparative advantage in violence: a tournament among western European
rulers that fostered military innovation. Politics made that tournament peculiar to
western Europe and led the continent to dominate the technology of artillery, firearms,
fortifications, and gunships.
The Evidence from Prices
Suppose that we confine ourselves to examining the cost of producing the new
weapons that played a key role in military revolution–artillery, handguns, and
gunpowder. The question would be whether the cost curves for producing these military
goods are declining, once we take into account changes in other prices. If the cost curves
are shifting down, then the production functions for the weapons are moving out, and the
firms producing them are undergoing technical change.
This sort of exercise certainly has its limits and is probably biased against finding
any technical change. To begin with, it likely to underestimate the magnitude of the
military revolution. Ideally, we should be measuring the cost of attaining a given level of
military effectiveness, but we are instead simply gauging the cost of producing certain
4
military products, and only doing that once the products are available for sale in
sufficient numbers to leave a historical record. Restricting our attention to the products
leaves out tactical innovations, better training, and improvements in provisioning armies
and navies and in raising money to pay for military operations. And by omitting
advances in ship construction, seaborn strategy, and maritime forces’ ability to fight
around the globe and in bad weather, it glosses over most of naval warfare, where
western Europe’s comparative advantage was probably greatest. Similarly, waiting until
prices appear in the historical records is likely to omit the initial drop in the cost of
producing the weapons right after they were first introduced but before sales and cost
estimates left much of trace in the archives.
In an ideal world, we could put together a long, homogenous series of prices for
artillery, handguns, and gunpowder in countries across the world. Unfortunately, we are
not at that stage yet, in large part because prices for military goods–guns in particular–are
hard to come by.
5
For the moment at least, we have to make do with somewhat
fragmentary price data from several western European countries only–in particular,
England, France, and (for a smaller number of observations) Germany.
What then do the price data for artillery, handguns, and gunpowder from these
countries tell us? Let us begin by assuming that each of these goods is each produced by
cost minimizing firms that are small relative to the size of the market they sell in and that
entry into these product markets is open. Let us also assume that markets for the factors
of production are competitive and that the firms have U-shaped short run average cost
curves.
6
These are not unreasonable assumptions for England, France, and Germany.
Factor markets were competitive, and weapons production in these countries was, for the
most part, in the hands of a large number of small scale contractors and independent
craftsmen. Furthermore, entry into the weapons business did seem to be open, at least in
the long run. Craftsmen and contractors moved their production from city to city and
even migrated from country to country. While there were some signs of fleeting
collusion or high prices in England and France when their rulers wanted to nurture the
native arms industry, they seem to have been temporary, because major weapons buyers
(this was true in particular of governments) would go elsewhere if they thought prices
were high.
7
Under these assumptions, it will be difficult for weapons producers to collude,
and free entry will drive them to produce at minimum average cost. The long run
industry supply curve will then be flat, and the cost of producing a quantity y of our
military good at time t will be turn out to be y c(w, t), where c(w, t) is the minimum
average cost of producing the good and w is the vector of factor prices. The function c(w,
t), which is also a firm’s marginal cost, will be independent of y but will depend on time
to allow for the possibility of technical change. If there is technical change, then c(w, t)
will be a decreasing function of t for any given w, and the partial derivative of its
logarithm will give the rate of technical change. (For technical details here and in what
follows, see the appendix.)
Because collusion will be difficult, the price p of the good produced will be the
marginal cost, or c(w, t). Provided that all of our assumptions held, we could therefore
test for technical change by regressing the price of each of our military goods on w and t.
All we would have to do is to choose a suitable functional form for c(w, t). Ideally, we
5
might want to use some flexible functional form, but lack of enough price observations
would probably limit us to deriving it from a Cobb-Douglas cost function, which would
at least be a first order approximation to c(w, t). The Cobb-Douglas technology will have
to constant returns to scale since the marginal cost is independent of output. If we adopt
the Cobb-Douglas functional form, and if the technology changes at a constant rate and is
cost neutral, then
ln (p) = ln (c(w, t)) = a - bt + s
0
ln (w
0)
+ . . . + s
n
ln (w
n)
+ u (1)
where a is a constant, b > 0 is the rate of technical change, u is an error term, s
i
and w
i
are
the factor share and price of the i-th factor of production, and the factor shares have to
add up to one. Equation 1 is equivalent to assuming that the good’s production function
is Cobb-Douglas with a multiplicative constant that grows at rate b. Because the factor
shares add up to one, we can single out one of the factor prices (say w
0
) and actually
estimate the following equation:
ln (p/w
0
) = a - bt + s
1
ln (w
1
/w
0
) + . . . + s
n
ln (w
n
/w
0
) + u (2)
where the only restrictions on the s
i
now are that they and their sum lie between zero and
one.
Unfortunately, we do not yet have enough data to do that, although it may become
possible in the future as more prices become available.
8
But if we let w
0
be the price of
skilled labor (an essential input into weapons production), then we can at least calculate
p/w
0
and compare how it changes with the variation in the relative prices w
1
/w
0
through
w
n
/w
0
. If p/w
0
, the relative price of military goods relative to skilled labor, falls more
rapidly than the relative prices of the other factors of production, then we have evidence
for technical change in the military sector, and we can estimate how large the rate of
technical change must have been.
If Figures 1 through 5 can be trusted, the price of military goods seems to have
fallen relative to the cost of skilled labor and relative to the cost of major factors of
production used in producing weapons in both England and France. Prices dropped for
artillery, muskets, and pistols, and they did so as early as late Middle Ages. Of course,
one might want to add a rental price of capital to the figures, but if we make reasonable
guess at depreciation and suppose that the sales price of capital goods was proportional to
skilled wages, then the rental price of capital declines only slightly in the figures, and if
the capital is building space, its rental price may have actually risen sharply, at least in
some locations (Figures 6 and 7). What the figures suggest, therefore, is that the military
sector of the economy witnessed sustained technical change over a long period of time
before the Industrial Revolution.
We can get a sense of how large the technical change must have been if we take
our earliest and latest price observations for each military good and use equation (2) to
estimate an upper bound for how much of the change in the price can be accounted for by
shifts in the costs of the factors of production. We know how much ln(p/w
o
) changed
between the first and last observation, and we know how much the terms ln(w
i
/w
0
)
changed too, at least for the factors of production listed in Table 1. Our coefficient b will
therefore equal
6
(–Δ
ln (p/w
0
) + s
1
Δ
ln (w
1
/w
0
) + . . . + s
n
Δ
ln (w
n
/w
0
) +
Δ
u
)/Δt
where Δ denotes the difference in each term between the initial and final period. This
expression will be greater than or equal to
(–Δ
ln (p/w
0
) + (1 – s
0
)
Δ
ln (w
i
/w
0
) +
Δ
u
)/Δt
where s
0
is the factor share of labor and Δ
ln (w
i
/w
0
)
is the smallest of the terms Δ
ln
(w
1
/w
0
)
, . . . , Δ
ln (w
n
/w
0
)
. If we take expectations (to make the Δu disappear) and
assume that the changes in the prices of the factors of production are all at least as large
as smallest one we can derive from Table 1, then we can calculate a lower bound for the
expected value of b simply by guessing at s
0
.
If we perform this calculation with a labor share of 0.5 (other reasonable labor
shares yield similar results), the resulting rates of technical change are nearly all larger by
preindustrial standards (Table 1). Apart from the 0.1 percent rate of change for French
muskets, the rates of growth in productivity are all over 0.5 percent per year, and the
figure is 0.9 percent for the manufacture of artillery in late medieval England. These
numbers compare favorably with rates of long run total factor productivity growth
elsewhere in the preindustrial world, which usually did not exceed 0.1 percent per year, at
least in sectors of the economy as large as the military one was in early modern Europe.
9
There were some exceptions to this rule–English agriculture, for instance, which seems to
have sustained long term total factor productivity growth rates of 0.2 to 0.3 percent per
year–but in most sectors of the preindustrial economy, faster growth could simply not be
sustained.
10
Even during the Industrial Revolution, total factor productivity growth in
Britain seems to have hovered between 0.1 percent per year and 0.35 percent per year.
11
How could the defense industry do so well over such long periods of time, and in two
economies–France and England–that for most of the years in the table were largely pre-
industrial?
One could of course argue that all the evidence here is a chance result, because it
all depends on initial and final price observations, which could vary randomly and be
buffeted about by the costs of factors of production that remain unobserved.
12
If we
enough data, we could settle the issue by estimating equation (2) and testing hypotheses
about the sign and magnitude of the coefficient b. But we cannot do that, even with
statistical methods that make up for missing data.
One thing we can do, however, is to compare the price of our military good with
that of a similar civilian commodity that involved a similar production process.
13
If the
civilian commodity was made with similar factors of production and similar factor
shares, and if the same economic assumption held for it too (small firms, open entry, U-
shaped short run average cost curves, competitive factor markets, and a Cobb-Douglas
production function), then equation 2 would apply to its price q too, and the logarithm of
p/q would be:
ln (p/q) = c - dt + e
1
ln (w
1
/w
0
) + . . . + e
n
ln (w
n
/w
0
) + v (3)
Here c is a constant, d is the rate of technical change for the military good minus that for
the non military good, v is an error term, and the e
i
’s are differences in the factor shares
for the two goods. If the factor shares for the two goods are nearly equal, then the e
i
’s
will be close to zero, and
7
ln (p/q) ≈ c - dt (4)
We could then regress ln (p/q) on time and come up with an estimate for d, the rate of
technical change for our non military good less that for our non military good. The
estimate will be biased because the variables ln (w
i
/w
0
) will be omitted from the
regression, but because the e
i
’s are small, the bias will be small too and may be either
positive or negative.
14
If production of the non-military good does not experience any
technical change, then d will be close to the rate of technical change b for the military
good. If there is technical change in production of the military good, the d we get from
equation (4) is likely to underestimate the rate at which the cost is declining. The key, of
course, will be finding non-military goods with factor shares similar to those of the
military goods–ideally, non-military goods whose production functions did not change.
This we can actually do, although we have to keep in mind that the coefficients
and estimated standard coefficient errors may be biased in an unknown way. In addition,
if we have prices of the factors of production for which the share differences e
i
are likely
to be relatively large, we can add them to the regression since they are likely to bias our
estimate of d the most.
15
The advantage of doing so is that we can find prices for factors
such as iron or capital, which may be used more intensively in either the military or
civilian good. We can include prices for these factors in a regression of ln (p/q) on a
constant and time and assume that the small e
i
’s for the other omitted variables will keep
their contribution to the bias small. That amounts to running regression (3) with some of
the ln (w
i
/w
0
) omitted, but it is possible to run such a regression when it would be
impossible to get enough data to run a regression with all the variables ln (w
i
/w
0
).
Table 2 shows what happens when we run either a regression based on equations
3 (with some missing variables) or equation 4. Again, the regressions involve the prices
of French and English handguns and artillery from the late Middle Ages to the eighteenth
century, and now gunpowder is included too. The prices of the English military goods
are expressed relative to the cost of spades, a non-military good that presumably had
factor shares roughly similar to those involved in the production of handguns, for like
spades, handguns were made of wood and metal. Admittedly, the factor shares were
probably different for artillery and gunpowder, and it no doubt took more metal to make a
firearm than a spade. But even cannons had wooden carriages, and wooden and metal
tools were used to manufacture gunpowder. Despite these disadvantages, though, using
the price of spades has certain virtues. Technical change in their production was
probably small before the eighteenth century, and there are repeated price observations
for spades with relatively little price variation at any given time. And where we have
enough data, we can compensate for the different factor shares for iron in military goods
by adding the relative price of iron to the regressions.